CVE-2026-57140: PraisonAI: unauthenticated AgentOS API exposes agents

CRITICALPoC AVAILABLECISA: TRACK*

Published September 15, 2026

CISO Take

PraisonAI's TypeScript AgentOS (versions 1.6.0 through 1.7.2) binds to 0.0.0.0 by default and exposes GET /api/agents and POST /api/chat with no authentication, so anyone who can reach the port can list agents and their instruction prefixes and invoke them. Because invoked agents can reach tools, memory, external APIs, credentials and workflow state, this is effectively remote control of an AI agent with whatever privileges it holds, which justifies the CVSS 9.4 rating. Exploitation is trivial (network, low complexity, no privileges), yet near-term likelihood looks moderate: EPSS is 0.64% (about the 51st percentile), it is not in CISA KEV, and there is no public exploit or Nuclei template, although CISA SSVC marks it TRACK*. Inventory every PraisonAI AgentOS deployment, upgrade past the initial 1.7.2 remediation (verify against the GHSA-9752-mhqh-h34f advisory for the fully fixed version), and until then restrict the listener to localhost or place it behind an authenticating reverse proxy or network ACL.

Sources:NVDEPSSGitHub AdvisoryATLAS

What is the risk?

High to critical for any internet- or network-reachable AgentOS instance. The flaw is a missing-authentication issue (CWE-306) with no user interaction or privileges required, and the default 0.0.0.0 bind makes accidental exposure likely, especially in containers and cloud VMs. Impact depends on the agent's tool and credential scope. Confidentiality and integrity impact are rated high, availability low. Exploitation signals are currently weak (EPSS 0.00642, no KEV listing, no public PoC), so urgency is driven by severity and ease of exploitation rather than observed attacks. The package risk score and dependent counts are unknown, so blast radius should be assessed per deployment. The vendor's note that 1.7.2 is only an 'initial remediation' means residual risk may remain after upgrading.

How does the attack unfold?

Discovery

Attacker finds a reachable AgentOS service bound to 0.0.0.0 and calls GET /api/agents without credentials to list agents and instruction prefixes.

AML.T0084

Initial Access

Attacker sends unauthenticated requests to POST /api/chat, exploiting the missing authentication middleware.

AML.T0049

Tool Invocation

Attacker directs a chosen agent to use its tools, memory and external APIs.

AML.T0053

Impact

Sensitive data and credentials are exfiltrated or workflow state is manipulated, leading to data loss and possible lateral movement.

AML.T0098

Discovery

Attacker finds a reachable AgentOS service bound to 0.0.0.0 and calls GET /api/agents without credentials to list agents and instruction prefixes.

AML.T0084

Initial Access

Attacker sends unauthenticated requests to POST /api/chat, exploiting the missing authentication middleware.

AML.T0049

Tool Invocation

Attacker directs a chosen agent to use its tools, memory and external APIs.

AML.T0053

Impact

Sensitive data and credentials are exfiltrated or workflow state is manipulated, leading to data loss and possible lateral movement.

AML.T0098

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
PraisonAI	pip	—	No patch
1 dependents61% patched~23d to patchFull package profile →

Do you use PraisonAI? You're affected.

How severe is it?

CVSS 3.1

9.4 / 10

EPSS

0.6%

chance of exploitation in 30 days

Higher than 51% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○Public PoC indexed (trickest/cve)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVNetwork

ACLow

PRNone

UINone

SUnchanged

CHigh

IHigh

ALow

What should I do?

1 step

1) Upgrade PraisonAI to a release containing the fix (1.7.2 or later) and confirm the status of the complete fix in advisory GHSA-9752-mhqh-h34f and the referenced commit 709a038. 2) Until patched, override the default host to 127.0.0.1 or a private interface and block the AgentOS port at the firewall, security group or Kubernetes network policy. 3) Put an authenticating reverse proxy (mTLS, OIDC or API-key gateway) in front of /api/agents and /api/chat. 4) Reduce agent privileges: remove unneeded tools, use scoped short-lived credentials, and require human approval for high-impact actions. 5) Detect: search access logs for unauthenticated requests to /api/agents and /api/chat from unexpected sources, scan internal and external ranges for exposed AgentOS listeners, and review agent tool-call and outbound API logs for anomalies since deployment. 6) Rotate any secrets accessible to exposed agents if exposure cannot be ruled out.

What does CISA's SSVC say?

DecisionTrack*

Exploitationnone

AutomatableYes

Technical Impacttotal

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Auth BypassData LeakageAgentAPIFrameworkAML.T0049- Exploit Public-Facing Application AML.T0053- AI Agent Tool Invocation AML.T0084- Discover AI Agent Configuration AML.T0085.001- AI Agent Tools AML.T0098- AI Agent Tool Credential Harvesting

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 15-Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6-AI system operation and monitoring

NIST AI RMF

MANAGE 2.3-Responses to previously unknown risks and mechanisms for disengaging or deactivating AI systems

OWASP LLM Top 10

LLM06-Excessive Agency

Frequently Asked Questions

What is CVE-2026-57140?

Is CVE-2026-57140 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2026-57140, increasing the risk of exploitation.

How to fix CVE-2026-57140?

1) Upgrade PraisonAI to a release containing the fix (1.7.2 or later) and confirm the status of the complete fix in advisory GHSA-9752-mhqh-h34f and the referenced commit 709a038. 2) Until patched, override the default host to 127.0.0.1 or a private interface and block the AgentOS port at the firewall, security group or Kubernetes network policy. 3) Put an authenticating reverse proxy (mTLS, OIDC or API-key gateway) in front of /api/agents and /api/chat. 4) Reduce agent privileges: remove unneeded tools, use scoped short-lived credentials, and require human approval for high-impact actions. 5) Detect: search access logs for unauthenticated requests to /api/agents and /api/chat from unexpected sources, scan internal and external ranges for exposed AgentOS listeners, and review agent tool-call and outbound API logs for anomalies since deployment. 6) Rotate any secrets accessible to exposed agents if exposure cannot be ruled out.

What systems are affected by CVE-2026-57140?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, multi-agent orchestration, agent tool-calling pipelines, agent memory and workflow state, LLM API gateways.

What is the CVSS score for CVE-2026-57140?

CVE-2026-57140 has a CVSS v3.1 base score of 9.4 (CRITICAL). The EPSS exploitation probability is 0.64%.

What is the AI security impact?

Affected AI Architectures

agent frameworksmulti-agent orchestrationagent tool-calling pipelinesagent memory and workflow stateLLM API gateways

MITRE ATLAS Techniques

AML.T0049Exploit Public-Facing Application

AML.T0053AI Agent Tool Invocation

AML.T0084Discover AI Agent Configuration

AML.T0085.001AI Agent Tools

AML.T0098AI Agent Tool Credential Harvesting

Compliance Controls Affected

EU AI Act: Art. 15

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE 2.3

OWASP LLM Top 10: LLM06

What are the technical details?

Original Advisory

PraisonAI is a multi-agent teams system. From 1.6.0 until 1.7.2, AgentOS in src/praisonai-ts/src/os/agentos.ts uses the 0.0.0.0 default from src/praisonai-ts/src/os/config.ts and registers GET /api/agents and POST /api/chat without authentication middleware. A remote caller who can reach the service can obtain agent names, roles, and instruction prefixes and can invoke a selected agent, potentially reaching its tools, memory, external APIs, credentials, and workflow state. An initial remediation was released in version 1.7.2.

Exploitation Scenario

An attacker scans for hosts exposing the AgentOS port and sends GET /api/agents without credentials, receiving agent names, roles and instruction prefixes. They identify an agent with database, email or cloud tools, then POST to /api/chat with crafted instructions that make the agent run those tools, dump data from memory or connected systems, and reveal API keys or tokens it holds. The attacker may then pivot using the harvested credentials or repeatedly call the agent to run up LLM costs.

Weaknesses (CWE)

CWE-306Missing Authentication for Critical FunctionalityPrimary CWE-306Missing Authentication for Critical Functionality

CWE-306 — Missing Authentication for Critical Functionality: The product does not perform any authentication for functionality that requires a provable user identity or consumes a significant amount of resources.

[Architecture and Design] Divide the software into anonymous, normal, privileged, and administrative areas. Identify which of these areas require a proven user identity, and use a centralized authentication capability. Identify all potential communication channels, or other means of interaction with the software, to ensure that all channels are appropriately protected, including those channels that are assumed to be accessible only by authorized parties. Developers sometimes perform authentication at the primary channel, but open up a secondary channel that is assumed to be private. For example, a login mechanism may be listening on one network port, but after successful authentication, it may open up a second port where it waits for the connection, but avoids authentication because it assumes that only the authenticated party will connect to the port. In general, if the software or protocol allows a single session or user state to persist across multiple connections or channels, authentication and appropriate
[Architecture and Design] For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.

Source: MITRE CWE corpus.